SDEGnO

Optimization and performance testing of CUDA-(multi)GPU-accelerated codes for the automatic parameterization of physical models.

SDEGnO (Stochastic Differential Equations on GPUs for Optimization) aims to develop high-performance, general-purpose simulators for the simulation and automatic calibration of physical models based on Stochastic Differential Equations (SDEs).

The initiative seeks to implement an integrated framework that leverages Monte Carlo techniques and global optimization algorithms (evolutionary strategies, swarm intelligence) to drastically reduce simulation time and energy consumption while ensuring high accuracy.

GitHub stars CUDA

High-Performance Computing Astrophysics
SDEGnO Logo

Key Features

  • Architectural Optimization: Maximizes computational efficiency using NVIDIA multi-GPU architectures, refined memory management, and SIMD (Single Instruction, Multiple Data) techniques.
  • Intelligent Calibration: Integrates advanced algorithms for the automatic search and calibration of physical parameters to dynamically adapt to various application scenarios.
  • Uncertainty & Sensitivity: Features novel algorithms designed to rigorously assess parameter uncertainty and system sensitivity.
  • Energy-Aware Computing: Drastically cuts down execution time and the corresponding energy footprint of massive stochastic simulations.

Project Phases & Funding Details

The SDEGnO project is structured to deliver a replicable, scalable framework providing strategic support to future HPC research groups.

Grant & Institutional Info
  • Grant: ICSC National Centre for HPC, Big Data and Quantum Computing (Spoke 3 – Astrophysics), funded by PNRR MUR – M4C2 – Investment 1.4.
  • CUP: C53C22000350006
  • Principal Investigator (UNIVE): Prof. Marco Salvatore Nobile
  • Research Team: Marco Salvatore Nobile, Sabina Rossi, Matteo Grazioso, Leone Bacciu
  • Duration: 01/09/2024 - 30/11/2025

Technical Highlights

CUDA & Multi-GPU

Fully leverages NVIDIA GPU architectures to accelerate complex Monte Carlo integrations of Stochastic Differential Equations.

Global Optimization

Utilizes Swarm Intelligence and Evolutionary Algorithms to automatically calibrate and tune physical models without manual intervention.

Astrophysics Core

Applied directly to predict cosmic radiation propagation within the heliosphere, turning chaotic high-energy data into insights.

View Code on GitHub Request Information

Core Publications

2026

  1. Massive stochastic simulation of cosmic rays propagation in the heliosphere: The COSMICA code
    Leone Bacciu, Matteo Grazioso, Giovanni Cavallotto, and 5 more authors
    Astronomy and Computing, 2026
    DOI Code
  2. The Role of Solar Modulation on Cosmic Ray Spectra at GV Rigidities
    Stefano Della Torre, Rachele Guidetti, Massimo Gervasi, and 7 more authors
    2026
    46th COSPAR Scientific Assembly, Florence, Italy
  3. Why Performance Matters: Accelerating Solar Modulation of Galactic Cosmic Rays with High-Performance Computing
    Leone Bacciu, Matteo Grazioso, Giovanni Cavallotto, and 5 more authors
    2026
    EGU (European Geosciences Union), Wien, Austria
    DOI
  4. Validation of COSMICA code for massive stochastic simulation of cosmic rays propagation in the heliosphere
    Stefano Della Torre, Leone Bacciu, Matteo Grazioso, and 5 more authors
    Astronomy and Computing, 2026
    DOI

2025

  1. COSMICA: a GPU-optimized code for solar modulation studies
    L Bacciu, G Cavallotto, S Della Torre, and 5 more authors
    2025
    DOI Code Slides
  2. Optimized solar modulation studies using COSMICA GPU-enhanced code
    Leone Bacciu, Giovanni Cavallotto, Stefano Della Torre, and 5 more authors
    2025
    ESWW2025 (European Space Weather Week), Umeå, Sweden